Analysis · Finance
A Better Harvest Forecast Does Not Prove a Better Forecasting System
By AWEI · AI-compiled · Published · Analysis prepared · 1 source · doi.org
Europe’s maize yield forecasts became more accurate as harvest approached: the median forecast error fell from about 12% early in the season to 8% one month before harvest. Yet improving within a season did not establish that the forecasting system was improving across years.
Marijn van der Velde and Luigi Nisini demonstrated this distinction in their evaluation of the European Union’s MARS crop yield forecasting system, first published online in 2018. Its archive contained 19,980 forecasts issued during 1993–2015. The system combined weather information, crop models, statistical analysis and analyst judgment.
For seven crops, the researchers compared early-season, pre-harvest and end-of-campaign forecasts with reported national yields during 2006–2015. They measured error size and tendencies to overestimate or underestimate. Separately, they fitted trends to annual median absolute percentage errors to test whether accuracy improved across years, using consistent country groups for each evaluation period.
Cereal forecasts generally improved during the growing season, while oilseed improvements were weaker. Across the EU-27 evaluation for 2006–2015, no crop showed a statistically significant improvement in final forecast accuracy over time. That is an absence of detected improvement under this test, rather than proof that every component or country stood still.
Even the reference data could move. The authors documented official yield statistics being revised years after publication, changing the yardstick for earlier forecasts. They suggested that better statistics might explain some longer-term gains, but did not establish that cause. The evaluation also excluded cultivated area, so it did not assess the complete production forecast.
For readers assessing market forecasts, the useful distinction is between getting closer to an outcome and demonstrating sustained forecasting progress. Compare estimates made at similar lead times, check which countries and categories are included, and identify the version of the outcome data. This study supports those evaluation questions; it does not establish price accuracy or financial returns.
Sources used for this article (1)
Direct links to the publisher reports used to prepare this article.
- Source 1
- Performance of the MARS-crop yield forecasting system for the European Union: Assessing accuracy, in-season, and year-to-year improvements from 1993 to 2015 Marijn van der Velde, Luigi Nisini
